Valve flow pressure control system

Abstract

A valve flow and pressure control system belongs to the field of valve control and comprises a flow regulating valve, a pressure regulating valve, a stop valve, a flow transmitter, a pressure transmitter and a controller. Stop valves are arranged on branches where the flow regulating valve and the pressure regulating valve are located; the flow transmitter is arranged on a branch where the flow adjusting valve is located and used for monitoring a flow signal. The pressure transmitter is arranged on a branch where the pressure regulating valve is located and used for monitoring a pressure signal; the controller adjusts the opening degree of the flow adjusting valve and the opening degree of the pressure adjusting valve according to the obtained flow signal and the obtained pressure signal, and when it is judged that the flow signal or the pressure signal exceeds the limit, the controller controls the cut-off valve to conduct cut-off control. According to the invention, the system structure is simplified while the control precision is improved, and the flow and pressure of the pipeline can be adjusted more directly and conveniently.

Description

Technical Field

[0001] The invention belongs to the field of valve control, and in particular relates to a valve flow pressure control system. Background Art

[0002] In the industrial production process, flow rate and pressure are important parameters in the process pipeline and need to be precisely controlled. Currently, the commonly used flow regulation schemes include direct regulation and throttling regulation. Direct regulation generally uses a flow control valve, which generates a control signal to drive the control valve through an intelligent PID control algorithm. However, due to the accuracy of the existing gas flow meter and the airflow disturbance before and after the valve, the control accuracy deviation is large and the adjustment range is limited.

[0003] In addition, direct regulation also has problems such as low control accuracy, complex equipment structure, and limited adjustment range. At the same time, there is a control conflict problem of mutual interference between flow and pressure regulation, and the lag of sensor signals will also affect the control accuracy and cause response delay problems. Summary of the invention

[0004] The present invention proposes a new type of valve flow pressure control system by adopting a decoupling control system, which can achieve precise control, high safety, energy saving and high efficiency, and high degree of automation. The system can establish a fluid dynamics model, use the relative gain matrix method to quantify the coupling degree, design a hierarchical control architecture, use dynamic matrix compensation feedforward decoupling at the bottom layer, use LSTM neural network prediction model and reinforcement learning optimizer at the AI ​​optimization layer, and combine the adaptive fuzzy PID decoupling algorithm, digital twin assisted optimization, multi-objective particle swarm optimization, online learning mechanism and other advanced technologies to achieve intelligent decoupling control.

[0005] The technical solution adopted by the present invention is: A valve flow pressure control system, comprising: A flow regulating valve, a pressure regulating valve, a shut-off valve, a flow transmitter, a pressure transmitter, and a controller. The flow regulating valve and the pressure regulating valve are respectively arranged on different branches of the same pipeline system, and the branches where the flow regulating valve and the pressure regulating valve are located are both provided with shut-off valves; The flow transmitter is arranged on the branch where the flow regulating valve is located, and is used to monitor the pipeline fluid flow in real time and transmit the flow signal to the controller in real time; The pressure transmitter is arranged on the branch where the pressure regulating valve is located, and is used to detect the pipeline pressure change in real time and transmit the pressure signal to the controller in real time; The controller is used to execute a control program including coupling degree judgment, building a control architecture, control algorithm execution and over-limit protection. Building the control architecture includes building a decoupling control layer and a dynamic optimization layer. The controller adjusts the opening of the flow regulating valve and the pressure regulating valve respectively according to the obtained flow signal and pressure signal. When the flow signal exceeds the maximum limit value, the controller controls the cut-off valve to cut off the channels on both sides of the pressure regulating valve through over-limit protection. When the pressure signal exceeds the maximum limit value, the controller controls the cut-off valve to cut off the channels on both sides of the flow regulating valve through over-limit protection.

[0006] The coupling degree judgment includes establishing a fluid mechanics model, calculating a relative gain matrix and determining the coupling degree.

[0007] The decoupling control layer obtains the transfer function matrix G(s) of the controlled object and then constructs a decoupling compensator. G(s)=[[G11(s), G12(s)], [G21(s), G22(s)]], Among them, G11(s) is the transfer function of the flow control valve controlling the flow channel, G22(s) is the transfer function of the pressure control valve controlling the pressure channel, and G12(s) and G21(s) are the cross-coupling channel transfer functions.

[0008] The dynamic optimization layer is optimized using an LSTM neural network prediction model and a reinforcement learning optimizer. The neural network prediction model inputs numerical values ​​of six dimensions for training and prediction.

[0009] The controller is also used to execute control accuracy judgment. The control accuracy judgment is achieved by collecting the data corresponding to the judgment standard after optimization by the dynamic optimization layer to determine whether the decoupling parameters after the AI ​​optimization layer meet the control accuracy requirements. If the accuracy requirements are met, the controller will import the decoupling parameters into the next step to execute the control algorithm. If the accuracy requirements are not met, the controller will re-import the decoupling parameters into the decoupling control layer.

[0010] The control algorithm execution includes executing an adaptive fuzzy PID decoupling algorithm, digital twin assisted optimization, multi-objective particle swarm optimization, and executing an online learning mechanism.

[0011] The over-limit protection determines whether the flow signal and the pressure signal exceed the maximum value by detecting the dynamic limit parameter.

[0012] The numerical values ​​of the six dimensions include pressure setting value, pressure actual value, flow setting value, flow actual value, temperature, and viscosity.

[0013] The construction of the reinforcement learning optimizer includes enhancing the state space, constraining the action space, and improving the reward function.

[0014] The beneficial effects of the present invention are as follows: the present invention improves the control accuracy by establishing a fluid dynamics model, using a relative gain matrix method to quantify the coupling degree, and designing a hierarchical control architecture, and effectively solves the problem of large control accuracy deviation caused by the accuracy of the gas flow meter and the disturbance of the airflow before and after the valve in the prior art. At the same time, the present invention simplifies the system structure, uses a regulating valve, a shut-off valve, etc. as an actuator, and does not require a complex multi-tube group to cooperate with a pressure reducing valve and a throttling orifice plate. Compared with the throttling mode adjustment, the system structure is simpler and can more directly and conveniently adjust the pipeline flow and pressure. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the hardware system structure distribution of the present invention.

[0016] Figure 2 It is a schematic diagram of the feedback control method of the present invention.

[0017] Figure 3 Schematic diagram of the control steps of the controller program of the present invention.

[0018] In the figure: 1. Flow regulating valve; 2. Pressure regulating valve; 3. Cut-off valve; 4. Flow transmitter; 5. Pressure transmitter; 6. Controller. DETAILED DESCRIPTION

[0019] The present invention is described in detail below with reference to the accompanying drawings.

[0020] like Figure 1 , Figure 2 A valve flow pressure control system includes a flow regulating valve 1, a pressure regulating valve 2, a cut-off valve 3, a flow transmitter 4, a pressure transmitter 5, and a controller 6. The cut-off valve 3 is respectively arranged on both sides of the flow regulating valve 1 and the pressure regulating valve 2. The flow transmitter 4 is arranged on both sides of the flow regulating valve 1 to monitor the pipeline fluid flow in real time and transmit the flow signal to the controller 6. The pressure transmitter 5 is arranged on both sides of the pressure regulating valve 2 to detect the pipeline pressure change in real time and transmit the pressure signal to the controller 6.

[0021] like Figure 3 , the controller 6 is used to execute a control program including steps such as coupling degree judgment, control architecture construction, control accuracy judgment, control algorithm execution and over-limit protection. The controller 6 adjusts the opening of the flow control valve 1 and the pressure control valve 2 respectively according to the obtained flow signal and pressure signal. When the flow signal or the pressure signal is judged to be over-limit, the controller 6 will control the cut-off valve 3 to cut off the control. When the flow signal exceeds the maximum limit value, the controller 6 will control the cut-off valve 3 to cut off the channels on both sides of the pressure control valve 2. When the pressure signal exceeds the maximum limit value, the controller 6 will control the cut-off valve 3 to cut off the channels on both sides of the flow control valve 1.

[0022] Step S1, coupling degree judgment. This step establishes a fluid dynamics model and uses the relative gain matrix method to quantify and judge the coupling degree between flow and pressure. The result of the coupling degree judgment layer will be input into the decoupling control layer to provide a basis for the decoupling control of the decoupling control layer. The execution of the coupling degree judgment includes the following steps: Step S101, establish a fluid dynamics model. That is, establish a fluid dynamics model according to the flow rate Q, pressure difference ΔP, valve coefficient Cv, medium density ρ, pipeline resistance coefficient ζ, and flow area Ap, wherein the flow rate Q and pressure difference ΔP are respectively expressed by specific models as follows: ,

[0023] Step S102, calculate the relative gain matrix. The relative gain matrix Λ is obtained by the following formula:

[0024] in, .

[0025] like , the partial derivative of the flow rate Q with respect to the valve opening μ1 of the flow control valve 1 That is Q11, the partial derivative of the pressure difference ΔP with respect to the valve opening μ1 of the pressure regulating valve 2 That is P11.

[0026] When λ is close to 1, it indicates good decoupling, and when it is close to 0, it indicates severe coupling. However, a single λ value cannot accurately reflect the degree of coupling of the system at this time, so a comprehensive judgment is required through the gain matrix Λ.

[0027] Step S103, determining the coupling degree. That is, determining the coupling degree between flow and pressure according to the relative gain matrix result. When any element in Λ is less than 0.2, it is determined that there is a serious coupling between flow and pressure.

[0028] Step S2, build a control architecture. The control architecture includes a decoupling control layer and a dynamic optimization layer. The bottom layer is the decoupling control layer, which uses dynamic matrix compensation feedforward decoupling and designs a decoupling compensator; the upper layer is the dynamic optimization layer, which uses an LSTM neural network prediction model and a reinforcement learning optimizer.

[0029] Step S201: Build a decoupling control layer and use dynamic matrix compensation feedforward decoupling. Obtain the controlled object transfer function matrix G(s) through the step response data obtained in step S1. G(s)=[[G11(s), G12(s)], [G21(s), G22(s)]] Among them, G11(s) is the transfer function of the flow channel controlled by flow control valve 1, and G22(s) is the transfer function of the pressure channel controlled by pressure control valve 2. G12(s) and G21(s) are the cross-coupling channel transfer functions.

[0030] Based on the technology of transfer function G(s), a decoupling compensator Gd(s) is designed.

[0031] Gd(s) = G(s)^(-1). The flow and pressure are initially decoupled by the decoupling compensator.

[0032] Step S202: Building an AI optimization layer, i.e., using an LSTM neural network prediction model, training and optimizing the prediction of values ​​with an input dimension of 6. These values ​​include pressure setting value, actual pressure value, flow setting value, actual flow value, temperature, and viscosity. Its hidden layer is 32 neurons.

[0033] The construction of a reinforcement learning optimizer involves augmenting the state space, constraining the action space, and improving the reward function.

[0034] The enhanced state space is defined as the state space S = {ep, eq, Δu}, where ep is the pressure deviation, eq is the flow deviation, and Δu is the control increment; Constraining the action space means setting the boundaries of the action space. The action space A = {Kp, Ki, Kd}, i.e., PID parameters, defines the intervals of Kp, Ki and Kd respectively through actual field data; Set the reward function R = w1*e^(-|ep|) + w2*e^(-|eq|) - w3*(Δu)², where w1, w2, w3 are weight coefficients, |ep| and |eq| are the absolute deviation values ​​of pressure and flow, respectively, and Δu is the control increment.

[0035] Step S3, control accuracy judgment. That is, judge whether the decoupling parameters after the AI ​​optimization layer meet the control accuracy requirements. If yes, return to step S2, if not, execute step S4. The control accuracy judgment requirements are: the pressure fluctuation amplitude is less than ±0.05MPa, and the flow tracking error is less than ±2%.

[0036] Step S4, control algorithm execution. The control algorithm execution includes executing the adaptive fuzzy PID decoupling algorithm, digital twin assisted optimization, multi-objective particle swarm optimization, and executing the online learning mechanism, and further optimizing the decoupling control algorithm after passing the accuracy requirements.

[0037] Step S401, execute the adaptive fuzzy PID decoupling algorithm, and realize adaptive regulation according to the fuzzy rule base and online parameter adjustment. The fuzzy rule base contains 49 rules, and the output increment is calculated by fuzzy reasoning. The online parameter adjustment is: Kp' = Kp + α*∂J / ∂Kp Ki' = Ki + α*∂J / ∂Ki Kd' = Kd + α*∂J / ∂Kd Where α is the learning rate and J is the performance indicator function.

[0038] Step S402: Execute digital twin assisted optimization. Digital twin assisted optimization is to establish real-time synchronization of a three-dimensional virtual model, and the model equation is: dX / dt = f(X, U) + DigitalTwin_Correction Where X is the system state variable, U is the control input variable, f(X, U) is the system dynamics equation, and DigitalTwin_Correction is the digital twin correction term. The data refresh rate is ≥100Hz.

[0039] Step S403: Execute multi-objective particle swarm optimization and perform convergence operation by setting the fitness function. Fitness function fitness = 0.7*ISE + 0.2*IAE + 0.1*control_effort, where ISE is the integrated square error and IAE is the integrated absolute error. When the fitness change Δf<0.01% lasts for 10 generations, the algorithm converges.

[0040] Step S404: Execute the online learning mechanism, collect data buffer (window time 200ms), and trigger incremental learning when it is detected that the KL divergence between the new data and the model exceeds the threshold of 0.05. Update the model in 5 epochs with a batch_size of 32.

[0041] Step S5: Over-limit protection. Over-limit protection is determined by real-time monitoring of dynamic amplitude limiting. Dynamic amplitude limiting ,in To optimize the output, and is the dynamic limiting range, the maximum limit ,in is the maximum allowable pressure difference, is the maximum allowable flow rate. When the maximum allowable differential pressure or the maximum allowable flow rate exceeds the setting, the system will adopt the over-limit safety protection mechanism. When the maximum allowable differential pressure exceeds the setting, the system will focus on the problem of the pressure difference exceeding the range. At this time, the cut-off valve 3 near the flow control valve 1 will cut off the flow control. Similarly, when the maximum allowable flow rate exceeds the set value, the system will focus on the flow problem. At this time, the cut-off valve 3 near the pressure control valve 2 will cut off the pressure control. When the system is in over-limit protection, it switches to the preset PID mode.

[0042] Through the above steps, the present invention realizes valve flow and pressure decoupling control with precise control, high safety, energy saving and high efficiency, and high degree of automation. Compared with the traditional feedforward decoupling method, the present invention does not require an accurate mechanism model, and has a higher computing speed. When encountering emergencies, the over-limit judgment mechanism of the present invention can also ensure that the flow or pressure is stabilized within the safe range in the first time, taking into account accuracy, speed and safety.

[0043] The present invention can be applied in many fields. In the field of industrial water supply and heating, the present invention decouples the flow rate and pressure at the same time to limit the maximum water supply while maintaining a constant pressure in the pipe network. In the field of chemical production, the present invention accurately controls the pressure of the reaction materials to avoid danger caused by overpressure in the pipeline. In the field of energy transportation, the present invention can balance the transportation efficiency and the pressure bearing capacity of the system. In the pharmaceutical and food fields, the present invention can be used to ensure the precise proportioning and pressure stability of fluids in a sterile environment.

[0044] The above content describes the specific features and implementation scheme of the present invention. It should be noted that the specific implementation of the present invention is not limited to the above-mentioned methods. As long as various non-substantial improvements are made using the technical solution of the present invention, or the concept and technical solution of the present invention are directly applied to other occasions without improvement, they are all within the scope of protection of the invention. The scope of protection required by the present invention is defined by the claims and their equivalents.

Claims

1. A valve flow pressure control system, characterized in that: include: A flow regulating valve, a pressure regulating valve, a shut-off valve, a flow transmitter, a pressure transmitter, and a controller. The flow regulating valve and the pressure regulating valve are respectively arranged on different branches of the same pipeline system, and the branches where the flow regulating valve and the pressure regulating valve are located are both provided with shut-off valves; The flow transmitter is arranged on the branch where the flow regulating valve is located, and is used to monitor the pipeline fluid flow in real time and transmit the flow signal to the controller in real time; The pressure transmitter is arranged on the branch where the pressure regulating valve is located, and is used to detect the pipeline pressure change in real time and transmit the pressure signal to the controller in real time; The controller is used to execute a control program including coupling degree judgment, building a control architecture, control algorithm execution and over-limit protection. Building the control architecture includes building a decoupling control layer and a dynamic optimization layer. The controller adjusts the opening of the flow regulating valve and the pressure regulating valve respectively according to the obtained flow signal and pressure signal. When the flow signal exceeds the maximum limit value, the controller controls the cut-off valve to cut off the channels on both sides of the pressure regulating valve through over-limit protection. When the pressure signal exceeds the maximum limit value, the controller controls the cut-off valve to cut off the channels on both sides of the flow regulating valve through over-limit protection.

2. A valve flow pressure control system according to claim 1, characterized in that: The coupling degree judgment includes establishing a fluid mechanics model, calculating a relative gain matrix and determining the coupling degree.

3. A valve flow pressure control system according to claim 1, characterized in that: The decoupling control layer obtains the transfer function matrix G(s) of the controlled object and then constructs a decoupling compensator. G(s)=[[G11(s), G12(s)], [G21(s), G22(s)]], Among them, G11(s) is the transfer function of the flow control valve controlling the flow channel, G22(s) is the transfer function of the pressure control valve controlling the pressure channel, and G12(s) and G21(s) are the cross-coupling channel transfer functions.

4. A valve flow pressure control system according to claim 1, characterized in that: The dynamic optimization layer is optimized using an LSTM neural network prediction model and a reinforcement learning optimizer. The neural network prediction model inputs numerical values ​​of six dimensions for training and prediction.

5. A valve flow pressure control system according to claim 1, characterized in that: The controller is also used to execute control accuracy judgment. The control accuracy judgment is achieved by collecting the data corresponding to the judgment standard after optimization by the dynamic optimization layer to determine whether the decoupling parameters after the AI ​​optimization layer meet the control accuracy requirements. If the accuracy requirements are met, the controller will import the decoupling parameters into the next step to execute the control algorithm. If the accuracy requirements are not met, the controller will re-import the decoupling parameters into the decoupling control layer.

6. A valve flow pressure control system according to claim 1, characterized in that: The control algorithm execution includes executing an adaptive fuzzy PID decoupling algorithm, digital twin assisted optimization, multi-objective particle swarm optimization, and executing an online learning mechanism.

7. A valve flow pressure control system according to claim 1, characterized in that: The over-limit protection determines whether the flow signal and the pressure signal exceed the maximum value by detecting the dynamic limit parameter.

8. A valve flow pressure control system according to claim 4, characterized in that: The numerical values ​​of the six dimensions include pressure setting value, pressure actual value, flow setting value, flow actual value, temperature, and viscosity.

9. A valve flow pressure control system according to claim 4, characterized in that: The construction of the reinforcement learning optimizer includes enhancing the state space, constraining the action space, and improving the reward function.

Images